WP/04/190

IMF Working Paper

The Internal Job Market of the IMF's

Economist Program
Greg Barron and Felix Vardy

I N T E R N A T I O N A L M O N E T A R Y FUND
 © 2004 International Monetary Fund WP/04/190

IMF Working Paper

IMF Institute

The Internal Job Market of the IMF's Economist Program

Prepared by Greg Barron and Felix Vardy1

Authorized for distribution by Sunil Sharma

October 2004

Abstract

This Working Paper should not be reported as representing the views of the IMF.
The views expressed in this Working Paper are those of the author(s) and do not necessarily represent
those of the IMF or IMF policy. Working Papers describe research in progress by the author(s) and are
published to elicit comments and to further debate.

This paper shows how the internal job market for participants in the IMF Economist Program
(EPs) could be redesigned to eliminate most of the shortcomings of the current system. The
new design is based on Gale and Shapley's (1962) Deferred Acceptance Algorithm (DAA)
and generates an efficient and stable outcome. An Excel-based computer program, EP-Match,
implements the algorithm and applies it to the internal job market for EPs. The program can
be downloaded from http://www.people.hbs.edu/gbarron/EP-Match_for_Excel.html.

JEL Classification Numbers: C78, D73

Keywords: Matching, stability, marriage problem, deferred acceptance algorithm,

Gale-Shapley, entry-level job market

Author(s) E-Mail Address: gbarron@hbs.edu and fvardy@imf.org

1
Greg Barron is an Assistant Professor at Harvard Business School and Felix Vardy is a
second-year participant in the Economist Program, currently in the Asian Division of the
IMF Institute. The authors would like to thank David Coe for encouragement to write this paper,
and Abdul Abiad, Chris Clarke, Andrew Feltenstein, Mohsin Khan, Jorge Marquez-Ruarte,
Dayalini Mendez, Benoit Mercereau, Miguel Messmacher, Jacques Miniane, Saleh Nsouli,
Al Roth, Martin Schindler, Sunil Sharma, Clinton Shiells, Jay Surti, and Zhiwei Zhang for
comments and suggestions. All errors are our own.
 - 2-

Contents Page

I. Introduction 3

II. The Current Situation 3

A. Decentralized Matching of Transferring and Graduating EPs 3
B. Centralized Matching of Incoming EPs 5

III. The Deferred Acceptance Algorithm 5

A. Preliminaries 5
B. The Algorithm 6
C. Strategic Issues 8
Strategy-proofness 8
The DAA and the Pressure to Pre-commit 9
Departments or Positions? 9
D. Summary 10

IV. The Computer Program 10

A. Description 10
B. Examples 11
Example 1 11
Example 2 14

V. Conclusion 16

Appendix: Matching of Incoming EPs 17

A. Introduction 17
B. Methodology 17
C. Stability 18
D. Pareto Efficiency 18
E. Conclusion 18

References 26

Figures:
1. EP Match: Instructions 20
2. EP Match: Departments' Preferences 21
3. EP Match: EPs' Preferences 22
4. EP Match: The Match 23

Tables:
1. Stable and Unstable Centralized Mechanisms 19
2. Participants' Preferences 24
3.HRD's Match 25
 -3-

I. INTRODUCTION

The Economist Program is the gate of entry into the IMF for young economists joining the
Fund soon after graduate school. Each year, 35 to 45 participants, usually referred to as
"EPs," join the Program for a two-year period. During this time, EPs work in two different
departments of the Fund. After graduating from the Program, EPs are considered for
permanent staff positions.

In the context of the Economist Program, each EP is matched to a department on three

occasions. First, when he joins the Fund in June or October;2 then, when he transfers to a
second-year position one year later; and, finally, when the EP is assigned to a permanent
position at the end of the Program.

Matching of EPs to second-year assignments and permanent positions takes place through a
decentralized, internal job market system. Matching of incoming EPs to first-year positions is
done centrally, by the IMF's Human Resources Department (HRD).

This paper argues that the procedures currently in place exhibit structural shortcomings. It
then proposes an alternative centralized matching mechanism, based on Gale and Shapley's
(1962) Deferred Acceptance Algorithm (DAA), that resolves most of these shortcomings.
Finally, it develops EP-Match, an Excel-based computer program that implements the DAA,
and applies it to the internal job market for EPs.

II. THE CURRENT SITUATION

A. Decentralized Matching of Transferring and Graduating EPs

Transferring and graduating EPs are matched to second-year assignments and permanent
positions through a decentralized job market system. However, to promote equality of
opportunities, and to prevent unraveling of the market,3 EPs and departments are supposed to
abide by strict rules as to when they can first contact each other, and when they can start
making and accepting job offers.

In this paper, the pronoun "he" refers to a person of unspecified gender.

3
By "unraveling" we mean the phenomenon, common in entry-level labor markets, that
employers make job offers earlier from year to year. This is done in an effort to preempt the
competition and recruit the best candidates. Once unraveling has started, it is hard to stop and
often spins out of control. In the market for federal appellate law clerks, for example, job
offers have recently been made almost two years in advance of employment. (See Avery and
others, 2001.) The same holds for entry-level gastroenterology positions for medical doctors.
Interviews for these positions now take place even before students have had the opportunity
to explore other subspecialties (see Niederle and Roth, 2003). In markets that experience
unraveling, applicants typically receive "exploding" offers that must be accepted or rejected
on short notice.
 -4-

In terms of Roth and Xing's (1994) taxonomy of entry-level labor markets, the internal job
market for transferring and graduating EPs is in "Stage 2." Here, "Stage 1" refers to an
entirely unregulated job market, while "Stage 3" refers to a fully centralized market. Their
detailed description of the symptoms of a typical "Stage 2" market is so recognizable that we
quote it here at some length.

"In stage 2, a (...) market organization (...) attempts to establish a uniform date before which offers
should not be made, and often an earlier date before which interviews should not be conducted, and a
later date (or time) before which candidates who have received offers should not be required to
respond. Sometimes this is hardly successful at all, with many market participants ignoring or
circumventing the rules, and those who obey them quickly finding that this puts them at a
disadvantage. And even when uniform dates are successfully established and maintained, the market
often experiences a great deal of congestion and chaotic behavior, as the deadlines for accepting or
rejecting offers grows near. A firm is eager to know if its offers will be accepted, in time so that if it
has unfilled positions it may approach its most preferred alternative candidates before they have had to
accept any offers they have received. And candidates who have received offers, but not from their first
choice firm, are intent upon waiting until the last allowable moment before accepting any offer, in the
hope of receiving a better one. Particularly if, as often seems to be the case, some fraction of
candidates holds on to multiple offers as the final deadline approaches, this means that just before the
deadline expires many transactions still remain to be made. Firms whose first choice candidates reject
them may now find that their next dozen candidates have already accepted offers, and candidates may
receive preferred offers moments after making a verbal commitment to accept an earlier offer. In some
markets such verbal commitments are virtually always honored, and in others they are sometimes
reneged on. In either event, in the aftermath, many firms and candidates have just missed making
connections they would have preferred. The result is that the following year witnesses a resurgence of
strategic behaviors designed to avoid being caught short at the end of the market. Often new rules are
formulated to prohibit the more brazen of these, and new adaptations are made. While some markets
have persisted for many seasons in this fashion, systems of formalized dates are often abandoned, with
the market either reverting to stage 1 [an unregulated market], or moving on to stage 3 [a centralized
market]." [Roth and Xing, 1994, p. 996]

From personal experience, one of the authors can attest that the internal EP job market for
transferring and graduating EPs exhibits virtually all of the symptoms described by Roth and
Xing. Specifically:

• The IMF's Human Resources Department (HRD) tries to prevent unraveling by

enforcing rules on market timing, e.g., timing of first contact, interviewing, making
and accepting offers.

• Departments and EPs routinely try to circumvent these rules, while those participants
who do obey the rules often end up with rather unfavorable outcomes. A prime
example is the rule that EPs and departments should not contact each other before a
certain date. This rule is clearly unenforceable.

• When a department makes a job offer to an EP, it often pressures the EP to accept or
reject the offer on the spot, or on very short notice.

• Conversely, if an EP can get away with it, he often tries to postpone a decision on a
particular offer until the very last moment, in the hope of receiving a better offer.
 -5-

• Sometimes, the hoarding of multiple offers by a few sought after EPs leads to a
virtual standstill in the market. And once these EPs have made up their minds, a
flurry of offers and acceptances takes place in a very short time period. (This kind of
"congestion" is studied extensively in Roth and Xing, 1997.)

• In response to continuous complaints, HRD has modified the rules governing the
market numerous times, apparently without much success.

In addition to the symptoms mentioned by Roth and Xing, it has been noted that:

• The current system hurts EPs who are traveling - or prevents departments from
sending EPs on a business trip during the job market period - because of the limited
time-frame in which EPs and departments are allowed to interact.

• The current system creates a lot of stress for the participants, as they consider the
process to be rather unpredictable and not particularly fair.

B. Centralized Matching of Incoming EPs

HRD matches incoming EPs to departments on the basis of preferences of EPs and
departments over each other. These preferences are solicited in the form of rank-order lists
(ROLs). Participants who do not respond are assumed to be indifferent.

The process by which HRD translates the participants' preferences into a match of EPs and
departments is heuristic. This means that HRD tries to find the "best fit" without formulating
an explicit objective function or a procedure to resolve conflicting preferences.

The drawback of a heuristic approach is that the resulting match is not guaranteed to be
Pareto efficient and/or stable.4 Indeed, we show in the Appendix that, for a recent cohort of
incoming EPs for which we were able to obtain the necessary data, HRD's match was in fact
neither.

Motivated by the shortcomings of the matching systems currently in place, in the next
section, we discuss Gale and Shapley's Deferred Acceptance Algorithm.

III. THE DEFERRED ACCEPTANCE ALGORITHM

A. Preliminaries

The internal job market for EPs can be viewed as a "College Admissions Problem" (CAP).
The CAP is concerned with two disjoint groups, which we shall refer to as EPs and
departments, that have to be matched to each other while taking into account their mutual

A formal definition of stability is given in the next section.

 -6-

preferences. The CAP was first formulated by David Gale and Lloyd Shapley in a path-
breaking paper in the American Mathematical Monthly (Gale and Shapley, 1962).

Gale and Shapley (GS) argue that a match of EPs and departments, denoted by m, solves the
CAP only if it is stable. That is, only if there is no (EP, department)-pair that would prefer to
be matched to each other instead of sticking to their matches prescribed by m. Obviously, a
matching process that violates this property can easily be upset by an EP and a department
getting together in a manner that benefits both. Such an (EP, department)-pair is called a
"blocking pair." The empirical literature indeed confirms that stability is a very important
property when designing a matching market to fix unraveling and associated "Stage 2"
market failures (see Table 1).

GS prove that, for the CAP, a stable match always exists.5 Although there may be multiple
stable matches for a given preference profile, when preferences are strict, the set of positions
filled and EPs employed is the same in all stable matches (Roth and Sotomayor, 1990).
Another attractive feature of stability in the CAP is that it implies Pareto efficiency. (This
follows from the fact that all stable matches are in the "Core." See, e.g., Roth and Sotomayor,
1990.) The reverse, however, is not true. That is, typically, there are many Pareto efficient
matches that are unstable. Thus, stability is a strictly stronger requirement than Pareto
efficiency.

In their original paper, GS also provide an iterative procedure for finding a stable match with
respect to the participants' stated preferences. This procedure is called the Deferred
Acceptance Algorithm. Currently, (a variant of) the DAA is used to match applicants and
employers in virtually all medical and dental residency programs, pharmacy practices, and
psychology internships, both in the United States and Canada.6 (See, e.g., www.nrmp.org.)

B. The Algorithm

The DAA is best illustrated in a simplified environment in which each department has
exactly one vacancy and the total number of vacancies is equal to the total number of EPs.
This simple set-up is known in the literature as the "Marriage Problem." Subsequently, we

5
That this result is far from trivial is illustrated by the closely related "Roommates Problem."
In the Roommates Problem, an even number of boys has to be paired-up amongst each other.
GS show that, in this case, a stable matching may not exist. Instead, cycles of blocking pairs
may cause the set of stable matchings to be empty: Consider four boys, A, B, C, and D. Let A
rank B first (i.e., B is A's most preferred roommate), while B ranks C first, and C ranks A
first. At the same time, A, B, and C all rank D last. Then, regardless of D's preferences, there
exists no stable matching; whoever is matched to D will want to move out, and one of the
other two will be willing to take him in.
6
The variant used is the Roth and Peranson algorithm (1999). Its advantage over the standard
DAA is that it can handle couples applying together.
 -7-

can modify the DAA to handle the general case with an arbitrary number of departments,
vacancies per department, and EPs.

It should be noted that, under the DAA, EPs and departments continue to make initial contact
and arrange interviews in a decentralized manner. But in the end, instead of making offers
directly, participants submit rank-order lists (ROLs) of each other to the market maker, i.e.,
HRD, who then runs the DAA. The output of the DAA is a stable match. That is, a stable
allocation of EPs over departments.

For ease of exposition, we describe the algorithm as if offers and rejections in the DAA are
made by the departments and EPs. In reality, it is the market maker that makes these
decisions for them, based on the ROLs submitted by the participants.

For the simple case ("Marriage Problem"), the DAA works as follows:

Round 0: EPs and Departments rank each other in order of attractiveness.

Round 1: Each department makes a job offer to its top-ranked EP. An EP who receives one
or more job offers rejects all but the most preferred among the offers received. The relatively
most preferred job offer is kept on hold.

Round i (i >2): A department that was rejected in the previous round makes an offer to its
top-ranked EP among those who have not yet rejected it. An EP who receives new offers
rejects all but the most preferred among the new offers received and the offer kept on hold
from the previous round. Again, the relatively most preferred offer is kept on hold.

Stop: The algorithm terminates when no new job offers are made. At that point each EP has
exactly one offer which he accepts.

Note that the roles of EPs and departments in the DAA can be reversed, such that it is the
EPs who make the offers to the departments. While both procedures lead to a stable match,
the outcomes are not necessarily identical. In fact, Gale and Shapley show that the DAA-
generated stable match is (weakly) preferred over all other stable matches by the side making
the offers, provided that preferences are strict.

The extension of the DAA to the general case is now straightforward. With k departments
each having a quota of qi positions, n EPs, and EPs making the offers, the algorithm proceeds
as follows. First, all EPs apply to the department of their first choice. Department i with qi
positions then places on its wait list the qi EPs whom it likes best, or all applying EPs if there
are fewer than qi . It rejects the rest. Rejected EPs then apply to the department of their second
choice. Again, each department selects the top qi among the new applicants and those on the
waitlist, puts these on its new wait list and rejects the rest. The procedure terminates when
every EP is either on a wait list or has been rejected by every department to which he is
willing to apply. At this point, each department accepts the offers from everyone on its wait
list and the EP-optimal stable match has been achieved.
 -8-

Though we have implicitly assumed that all departments rank all EPs (and vice versa), in
practice this is not really necessary. Instead, for each of its positions, a department should
rank sufficiently many EPs such that it is reasonably assured of at least one EP accepting the
job.7 This implies that the effort spent on interviewing and ranking EPs under the DAA will
not differ much from the effort expended under the current system. Also in the current
system, a department will want to avoid a situation in which, as a result of too many
rejections, it has to start interviewing additional EPs after the market has opened and offers
have been made. Because, by then, very few EPs are left to choose from.

Finally, rankings do not have to be strict. That is, being indifferent between some or all
counterparts is allowed and resolved by randomization.8 Also, departments can classify
certain EPs as unacceptable, and EPs can do the same for departments. We will get back to
these issues later.

C. Strategic Issues

Strategy-proofness
Roth (1982) proves that there exists no stable matching mechanism, including the DAA, that
is one-hundred-percent strategy-proof. That is, even for the DAA, one can "cook-up" a
constellation of preferences such that at least one participant would gain by distorting his true
preferences.

The catch is that strategic misrepresentation tends to require much more information about
other participants' preferences than is usually available. When participants in the DAA are
sufficiently uncertain about the preferences of others, Roth and Rothblum (1999) show that
they cannot gain by reversing the order of their true preferences. However, if towards the
bottom of his ROL a participant is close to indifferent between being matched or not, and the

7
For example, in the National Resident Matching Program, each employer interviews and
ranks only a tiny fraction of all potential candidates. This does not interfere with the
workings of the DAA.
8
When a large number of participants on both sides of the market are indifferent between
many counterparts, a complication may arise. For example, suppose that some EPs and some
departments do not care who they are matched to, or have not submitted a ROL. In that case,
the DAA proceeds with randomly generated (strict) preferences. Now, what if on the basis of
these randomly generated preferences, an "indifferent" EP and an "indifferent" department
end up matched to each other? Obviously, neither of them can or will object to that.
However, there may be other, "non-indifferent" EPs or departments that would very much
like to be matched to one of these "indifferent" participants. In that case, we might be able to
accommodate one of them without hurting anybody else. In other words, when the
randomization is such that "indifferent" participants end up matched to each other, the
matching may not be strictly Pareto optimal. In that case, we should either redo the DAA
with newly generated random preferences, or "manually" modify the matching to restore
strict Pareto optimality. See the Appendix for a real-world example.
 -9-

other side makes the offers, a participant might gain by cutting his ROL short, falsely
claiming that the least preferred options are "unacceptable." But in reality such a strategy is
unlikely to be used much, because of its high risk: it may cause the EP or department to
remain unmatched involuntarily.

In any case, truncation clearly is not an option for EPs, unless they are prepared to quit and
leave the Fund. For departments, truncation is only possible in the market for permanent
positions, since, at the end of the day, all transferring EPs have to be matched to a
department.

Roth and Rothblum's result of "almost-strategy-proofness in practice" goes a long way in

explaining the popularity and robustness of the DAA in real-world applications such as the
National Residency Matching Program for physicians. Indeed, Roth (1984, 1990, 1991) and
Roth and Xing (1994) show that, with very few exceptions, centralized matching
mechanisms survive if, and only if, they produce stable matches with respect to the
participants (stated) preferences. This is illustrated in Table 1, taken from Roth and
Rothblum, 1999.

The DAA and the Pressure to Pre-commit

In the current system, departments and EPs routinely flout the rules by committing to each
other before the official opening of the market. This raises the question why the situation
would be any different under the DAA. Specifically, what prevents departments and EPs
from reducing the DAA to a mere formality by pre-committing to top-rank each other?

In the current system, having an offer rejected in the official market is potentially very costly
for a department. The reason is that a rejected department may find that all other attractive
EPs have already accepted offers from other departments. To mitigate this risk, departments
routinely put extreme pressure on EPs to pre-commit to accepting a forthcoming job offer.

By contrast, rejections are costless in the DAA (at least, from a strategic perspective). The
reason is that all acceptances and rejections occur simultaneously. Therefore, a department
has no particular interest in making sure that its job offer is accepted by the first EP it is
offered to. Indeed, the zero rejection cost in the DAA implies that a department can and
should top-rank its favorite EP even if the probability of the EP accepting the offer is very
small. In turn, this undermines the credibility of the exploding nature of a pre-market offer,
as the department will top-rank its favorite EP, even if he refuses to pre-commit.

These considerations explain why, over time, unraveling and pre-commitment tend to occur
less under the DAA than in decentralized markets. Nevertheless, it is true that pre-
commitment might still occur on a smaller scale, particularly with repeated interaction. In
this context, Niederle and Roth (2004) stress the importance of "market culture," such as
norms governing exploding offers.

Departments or Positions?
So far, we have assumed that EPs rank, and are matched to, departments. Usually, however,
EPs do not have preferences over departments as such, but over specific positions within
 -10-

departments. For example, while an EP's first choice might be to work on the China desk in
the Asia and Pacific Department (APD), he might very well prefer to work on South Africa
in the African Department over working on Tuvalu in APD. In this case, APD can do two
things: either list the China and Tuvalu jobs separately, thereby allowing EPs to express
interest specifically in China or specifically in Tuvalu; or lump both jobs together, only
allowing for expressions of interest in working for APD without formal assurance of being
assigned to any job in particular.

When EPs have preferences over particular positions instead of over departments, there is a
trade-off between ex-ante clarity for EPs and ex-post flexibility for departments. If a
department chooses to specify its vacancies in great detail, it lowers the risk and/or ambiguity
for the applicants. In turn, this may help the department attract better EPs. On the other hand,
if a department chooses to be less specific, it retains full flexibility to assign EPs to one job
or another, depending on the need of the moment.

Presumably, competition will force all but the most popular departments to list their
vacancies separately. From a transparency perspective, separate listings are indeed
preferable.

D. Summary

The DAA solves many of the problems commonly encountered in matching markets. In the
DAA, all final acceptances occur simultaneously. This eliminates premature decisions based
on incomplete information. Also, undesirable phenomena such as unraveling, hoarding of
multiple offers, and reneging on a prior acceptance will no longer occur. At the same time, it
is no longer necessary to impose rules on the participants that are hard to enforce. In
particular, EPs and departments can communicate with each other as early and as often as
they wish. The only hard constraint is that participants hand in their rankings of each other on
a pre-announced date. The submitted rank-order lists (ROLs) then determine the matching of
EPs to departments (or positions) through application of the DAA. Operationally, a market
maker running the algorithm can execute all steps of the DAA in "real-time," one after the
other, provided that EPs and departments have handed in their ROLs. This eliminates the
need for repeated communication between EPs and departments, and thereby the need for
EPs to be in Washington during any particular time period.

IV. THE COMPUTER PROGRAM

A. Description

To facilitate implementation of the DAA at the Fund, we have developed an Excel-based

computer program, called EP-Match, that executes the algorithm and applies it to the Fund's
internal job market for EPs. The user-interface of the program consists of four spreadsheets:

1. "Instructions" (see Figure 1).

2. "Departments' Preferences" (see Figure 2).

 - 11 -

3. "EPs' Preferences" (see Figure 3).

4. "Match" (see Figure 4).

To find the stable match generated by the DAA, the user, i.e., the market maker, sequentially
works his way through the four spreadsheets. The first spreadsheet gives instructions on how
to use the program and requires the market maker to enter the number of EPs and the number
of departments (or positions) participating in the matching process. Here, the market maker
must also indicate whether he wants to run the EP-proposing or department-proposing variant
of the DAA. In the second spreadsheet, the market maker specifies the number of vacancies
in each department and enters the departments' rank-order lists (ROLs) in a matrix. The EPs'
ROLs go into the third spreadsheet. Finally, the fourth and last spreadsheet presents the
resulting stable match.

The program is quite versatile. First, it allows for an arbitrary number of EPs, departments,
and positions per department. Second, departments are allowed to be indifferent between
certain EPs, and EPs between positions. Third, a department can be assured it will not be
matched to an EP it finds unacceptable by not ranking that EP at all. Of course, it should only
do this if it prefers the position to remain unfilled over hiring that EP. In principle, EPs can
do the same with departments they do not want to join under any circumstance.

Finally, the program can also handle the following, slightly more complex situation. Suppose
department Dx has multiple positions on offer, but its ranking of EPs is not the same for all
positions. This might happen if a certain position requires particular skills, such as fluency in
a foreign language or financial expertise. In that case, we partition the set of Dx's positions
into r disjoint sub-setsS1, ..., Sr, such that, for all positions in a particular sub-set, Dx's
ranking of EPs is in fact the same. Then, department Dx is replaced by multiple "virtual
departments" Dx1, ..., Dxr, such that Dxi offers positions Si and ranks the EPs accordingly.
Finally, the program is run on the modified matching problem.

B. Examples

To illustrate how EP-Match works in practice, we now go through two examples.

Example 1
There are 3 departments, Dl, D2, D3, and 6 EPs, EP1, ..., EP6. Each department has 2
vacancies. Let the departments' ROLs be as follows:
 -12-

Departments' Rankings of EPs

Dl D2 D3
EP3 EP1 EP3
EP1 EP3 EP1
EP4 EP4 EP4
th
EP6 EP5 EP2
c th
EP2 EP2 EP6
EP5 EP6 EP5

This means, for example, that Dl likes EP3 best, EP1 second-best, and so on.

Similarly, the EPs' ROLs are:

EPs' Rankings of Departments

EP1
EPl | IEP2 EP3 EP4 EP5 EP6
st
1 Dl Dl D2 D3 Dl Dl
D2 D3 Dl Dl D3 D3
D3 D2 D3 D2 D2 D2

This means, for example, that EP4 has a preference ordering of departments given by D3,
D1,D2.

To solve for the associated department-optimal stable match using EP-Match, the market
maker first enters the number of departments (3) and EPs (6) in the appropriate boxes on the
"Instructions" spreadsheet, and then checks that the option "Dept. Proposing" is selected. On
the "Departments' Preferences" spreadsheet, the departments' rankings of the EPs is coded
by assigning rank-order " 1 " to the most-preferred EP, rank-order "2" to the next-most-
preferred EP, and so on. On the same spreadsheet, the number of vacancies ("positions") in
each department (i.e., 2, 2, 2) is entered. The departments' preference matrix then looks like:
 -13-

Clear All
D1 D2 D3
Positions 2 2 2
EP1 2 1 2
EP2 5 5 4
EP3 1 2 1
EP4 3 3 3
EP5 6 4 6
EP6 4 6 5

Similarly, on the "EPs' Preferences" spreadsheet, the market maker codes the EPs' rankings
of the departments by assigning rank-order " 1 " to the most preferred department, rank-order
"2" to the next-most preferred, and rank-order "3" to the least preferred department. The
EPs' preference matrix then looks like:

Clear All
EP1 EP2 EP3 EP4 EP5 EP6
D1 1 1 2 2 1 1
D2 2 3 1 3 3 3
D3 3 2 3 1 2 2

Finally, the "Match" button on the "Match" spreadsheet can be pressed, and the department-
optimal stable match appears on screen as follows:

Match
I
EP Department Department EP
EP1 D1 D1 EP1
EP2 D3 D1 EP6
EP3 D2 D2 EP3
EP4 D3 D2 EP5
EP5 D2 D3 EP2
EP6 D1 D3 EP4

Here, the list on the left is ordered by EP, while the list on the right is ordered by department.
Note that stability implies that no department will be able to lure a more preferred EP from
another department. Neither can any EP succeed in "stealing" a more preferred position from
another EP.

To calculate the EP-optimal stable match, select the option "EP Proposing" on the
"Instructions" spreadsheet and press the "Match" button on the "Match" spreadsheet. It is
easily verified that, in this example, the EP-optimal and departmental-optimal stable matches
are identical.
 - 14-

Example 2
In our second example we illustrate how EP-Match can be used to solve more complicated
scenarios, when not all EPs are acceptable to all departments, when there are more (or less)
positions than EPs, and when departments are indifferent between some EPs, and EPs are
indifferent between some departments.

Let us take the set-up of Example 1 as our point of departure but assume that department Dl
is indifferent between EP1 and EP4, and also between EP2 and EP5. Moreover, department
D2 now considers EP2, EP5, and EP6 to be unacceptable, while D3 has two additional
vacancies (i.e., 4 instead of 2). On the EP side, EP2, EP4 and EP6 are assumed to be
indifferent between the department they have ranked second and the department they have
ranked third. For the rest, the situation is the same as in Example 1.

In EP-Match, on-screen instructions explain how indifference and unacceptability can be

coded into the preference matrices. To express that department Di is in fact indifferent
between EPx and the EP ranked one lower than EPx, add a "*" to the rank-order number of
EPx on the "Departments' Preferences" spreadsheet. In our example, this means that Dl's
rank-order numbers of EP1 and EP2, i.e., "2" and "5," respectively, get a "*". To express that
department Di considers EPy unacceptable, leave cell (EPy, Di) empty. In our example, this
means that cells (EP2, D2), (EP5, D2), and (EP6, D2) remain empty.
 - 15-

The departments' preference matrix then looks like:

Clear All
D1 D2 D3
Positions 2 2 4
EP1 2* 1 2
EP2 5* 4
EP3 1 2 1
EP4 3 3 3
EP5 6 6
EP6 4 5

Incorporating EP2, EP4, and EP6's indifference between their second and third-ranked
departments, we get for the EPs' preference matrix:

Clear All
EP1 EP2 EP3 EP4 EP5 EP6
D1 1 1 2 2* 1 1
D2 2 3 1 3 3 3
D3 3 2* 3 1 2 2*

Now, the department proposing stable match is:

Match
I
EP Department Department EP
EP1 D1 D1 EP1
EP2 D3 D1 EP6
EP3 D2 D2 EP3
EP4 D3 D2 — not matched —
EP5 D3 D3 EP2
EP6 D1 D3 EP4
D3 EP5
D3 — not matched —

Again, the list on the left is ordered by EP, and the list on the right is ordered by department.

This stable match differs little from the stable match of Example 1. EP5 has moved from D2
to D3, filling one of D3's two additional vacancies, while D2's second and D3's fourth
vacancy go unfilled. The latter reflects the fact that, in our example, there are more positions
than EPs. The EP-proposing stable match is again identical to the department-proposing
stable match.
 -16-

V. CONCLUSION

In this note we have argued that adopting the Deferred Acceptance Algorithm to match EPs
and departments can enhance the well-being of all parties involved. Unlike the current
system, the DAA generates an efficient and stable outcome by optimally using all available
information on the preferences of the participants as contained in their rank-order lists. In
addition to its attractive theoretical properties, the DAA has been extensively tried-and-tested
in practice, and it was shown to produce excellent results. To facilitate implementation of the
DAA at the Fund, we have developed an Excel-based computer program that executes the
algorithm on the basis of EPs' and departments' rankings of each other.
 -17- APPENDIX I

MATCHING OF INCOMING E P S

A. Introduction

As described in Section II, HRD matches incoming EPs to first-year positions on the basis of
EPs' and departments' preferences over each other. The question that arises is whether these
matches are stable and/or Pareto efficient. Because HRD uses rules of thumb to do the
matching (instead of an algorithm), this question can only be addressed by looking at actual
data.

In this appendix we show that HRD's match was neither stable nor Pareto efficient in the
case of a cohort of incoming EPs for which we were able to retrieve the necessary data.
These data consist of the participants' preferences, given by their rank-order lists (ROLs)
(Table 2), and the match implemented by HRD (Table 3).9

B. Methodology

Participants (i.e., departments and EPs) who chose not to submit a ROL are assumed to be
indifferent as to who they are matched to. Similarly, participants who ranked only a subset of
counterparts are assumed to be indifferent between those left unranked, while strictly
preferring ranked over unranked counterparts.

To check the stability of HRD's match, one might be tempted to run EP-Match on the
participants' preferences and compare the outcomes. However, this would be misguided,
because the set of stable matches usually contains many elements while, in principle, EP-
Match only finds two of them; namely, the department-proposing and the EP-proposing
stable matches. Thus, discrepancies between HRD's match and those calculated by EP-
Match are not necessarily informative.

A second complication arises from the large number of declared and imputed "indifferences"
in participants' preferences. Only when preferences are strict does stability imply Pareto
efficiency. Else, stable matches may contain "suspect pairs." A suspect pair is a match of an
EP and a department, such that both the EP and the department are indifferent between
remaining matched to each other, or being matched to someone else. If the EP and/or the
department making up the suspect pair are very sought after among other departments or
EPs, it may be possible to make one or more of these other departments or EPs better off by
breaking up the suspect pair, without hurting anybody else. This implies that, with a lot of
indifference on both sides of the market, a stable match is only a candidate solution. To be a
full-fledged solution, it must be checked that none of the suspect pairs, if any, give rise to
Pareto inefficiencies.

9
To protect confidentiality we have made certain changes to the data. In particular, "dummy"
EPs, departments, and preferences have been added. However, the gist of the example and
the validity of the conclusions are fully retained.
 - 18- APPENDIX I

Therefore, a two-step procedure is followed to assess the optimality of HRD's match with
respect to the participants' preferences:

• Check the stability of the match by looking for blocking pairs.

• Check whether suspect pairs, if any, give rise to Pareto inefficiencies.

C. Stability

Is HRD's match stable with respect to the participants' stated preferences? The answer is no.
To see why, note that EP7 and department Dl form a blocking pair: EP7 strictly prefers Dl
over his current match D5, and Dl strictly prefers EP7 over its current match EP2.

In this case, a simple exchange is sufficient to transform HRD's match into a stable one. That
is, we reassign EP7 to Dl and EP2 to D5, while keeping everything else the same. Both EP7
and D5 go from being matched to a counterpart who was not even on their submitted
preference list to being matched to their 1st choice. On the other hand, EP2 goes from his 1st
choice to his 2nd choice, while D5 goes from its 3rd choice to its 6th choice.

D. Pareto Efficiency

It is easily checked that HRD's match as given in Table 3 contains four suspect pairs,
namely, (EP3, D12), (EP10, D13), (EP13, D8), and (EP20, D8). Not all of these suspect pairs
give rise to Pareto inefficiencies, however. Only the first two suspect pairs do. By breaking
up (EP3, D12) we can make EP5 strictly better off without hurting anyone else, while
breaking up (EP10, D13) allows us to make EP18 strictly better off.

Specifically, we perform the following Pareto improving reassignments:

l.(EP3,D12) (EP3, D l l )
(EP5, Dl1) (EP5, D12)

2. (EP10, D13) • (EP10, Dl1)

(EP18,D11) (EP18, D13)

In this case, EP5 achieves his 1st choice instead of his 2nd choice, while EP18 achieves his 1st
choice instead of being assigned to a department not on his preference list. At the same time,
none of the others is worse off than before.

E. Conclusion

We have shown that HRD's match of incoming EPs to departments was neither stable nor
Pareto efficient. Application of EP-match followed by a critical assessment of all suspect
pairs would have resulted in a better outcome.
 -19-

Table 1
STABLE AND UNSTABLE CENTRALIZED MECHANISMS

Market Stable/Based on DA Algorithm • • Still in U s e

Entry level medical markets:

US. (NRMP) . yes/yes yes
Edinburgh ('69) yes/yes yes
Cardiff yes/yes yes
Canada yes/yes yes
Cambridge no yes
London Hospital no yes
Birmingham no no
Edinburgh ('67) no no
Newcastle no no
Sheffeld no no
Other markets:
Medical Specialties
(approximately 30 markets) yes/yes yes
Canadian Lawyers (articling positions)
Toronto yes/yes yes
Vancouver yes/yes no (abandoned in '96)
Calgary & Edmonton yes/yes yes
Dental Residencies (5 specialties,
2 general programs) yes/yes yes
Osteopaths (<'94) no no
Osteopaths (>'94) yes/yes yes
Pharmacists yes/yes yes
Sororities yes (at equilibrium)/ho

(Source: Roth and Rothblum, 1999.)

 -20-

Figure 1

Instructions:
for Excel
1. Enter the number of Departments and EPs in the box below.

Number of Departments: 3
Number of EPs: 6

Dept. Proposing C
EP Proposing E

2. Fill out the "Departments' preferences" worksheet according to the instructions on the sheet.
You can customise the Departments' and EPs' labels.
3. Indicate the number of open positions for each firm on the "Departments' preferences" worksheet.
4. Fill out the "EPs' preferences" worksheets according to the instructions on the sheet.
Here too, you can customise the Departments' and EPs' labels. n
5. Click the [Match] button. •
The Department-optimal stable match will appear in the "Match" spreadsheet. Save or print your
results!
Do not add/remove rows or columns from these 4 worksheets or change their order.

Created by: Greg Barron - 11/2003

CLER Research fellow - Research Computing Services
Harvard Business School
Thanks to Prof. A\ Roth for valuable advice. Any remaining errors are the author's. 11
The deferred acceptance algorithm: David Gale and Lloyd Shapley (1962).
 -21-

Figure 2
Departments' preferences:
- Each column contains the preference list of one department.
- You can change the names of departments and EPs.
- For each column, enter " 1 " in the row of the most preferred EP,
"2" in the row of the 2nd-most preferred EP and so on.
- Add a "*" to EP x's rank-order to express a department's
indifference between EP x and the EP y ranked one lower than EP x.
E.g., "3*" and "4*" in the same column means that the department
is indifferent between the EPs (s)he has ranked 3rd, 4th and 5th.
- Empty cells represent unacceptable EPs.
- Enter the number of positions being offered by each department
in the "positions"-row.

- Do not add/delete rows or columns from this sheet.

Clear All
Department 1 Department 2 Department 3 Department 4
Positions 2 2 2
EP 1 2* 1 2
EP 2 5* 5 4
EP 3 1 2 1
EP 4 3 3 3
EP 5 6 4 6
EP 6 4 6 5
EP 7
EP 8
EP 9
 -22-

Figure 3
EPs' preferences:
- Each column contains the preference list of one EP.
- You can change the names of departments and EPs.
- For each column, enter " 1 " for the most preferred
department, "2" for the 2nd-most preferred department and
so on.
- Add a "*" to department x's rank-order to express an EP's
indifference between department x and the department y
ranked one lower than department x.
E.g., " 1 * " means that the EP is in fact indifferent between
the department (s)he has ranked 1st and the department (s)he
has ranked 2nd.
- Empty cells represent unacceptable departments.
- Do not add/delete rows or columns from this sheet.

Clear All
EP 1 EP 2 EP 3 EP 4 EP 5 EP 6 EP 7 EP 8
Department 1 1 1 2 2 1 1
Department 2 2 3 1 3 3 3
Department 3 3 2 3 1 2 2
Department 4
Department 5
Department 6
 -23-

Figure 4
The Match:
- Click the [Match] button below.
- The left list is ordered by EP, and the right list is ordered by Department

Match

EP Department Department EP
EP 1 Department 1 Department 1 EP 1
EP 2 Department 3 Department 1 EP 6
EP 3 Department 2 Department 2 EP 3
EP 4 Department 3 Department 2 EP 5
EP 5 Department 2 Department 3 EP 2
EP 6 Department 1 Department 3 EP 4
 -24-

Table 2
EPs' Preferences over Departments Departments' Preferences over EPs

EP Participant Preferences 1/ Department Preferences 1/

EP1 D12, Dl1*, D13*, D5*, D7*, D4 Dl EP7,EP15,EP17

EP2 Dl, Dl1*, D13*, D5*, D7*, D12*, D4 D2 EP12, EP7, EP25
EP3 No preference indicated D3 EP19,EP22*,EP11*,EP16
EP4 D1,D6,D13,D12 D4 EP22, EP19, EP6, EP11, EP12
EP5 D12,D11 D5 EP15, EP21, EP7, EP6, EP19, EP2
EP6 No preference indicated D6 EP15, EP1, EP26
EP7 Dl D7 EP17,EP14,EP25
EP8 No preference indicated D8 No preference indicated
EP9 No preference indicated D9 No preference indicated
EP10 No preference indicated D10 EP9*,EP19
EP11 No preference indicated Dll EP22, EP6, EP19, EP21, EP23, EP7
EP12 D1,D5,D13 D12 No preference indicated
EP13 No preference indicated D13 No preference indicated
EP14 No preference indicated D14 E15,EP24
EP15 D1,D2,D12
EP16 Dl, D3, D10, D13,D5 1/ Preferences listed in ranked order
EP17 No preference indicated * Indicates indifference between EPX and the EP ranked one
EP18 D13,D1 lower than EPx.
EP19 No preference indicated
EP20 No preference indicated
EP21 D1,D5,D12,D13
EP22 No preference indicated
EP23 D5
EP24 D14, Dl
EP25 No preference indicated
EP26 D1,D6,D2

1/ Preferences listed in ranked order

* Indicates indifference between department Di and the department
ranked one lower than D i .
 -25-

Table 3
EPs' Departmental Assignments Departments' EP Assignments

EP Assigned Department Department Assigned EP

EP1 D6 Dl EP2
EP2 Dl Dl EP15
EP3 D12 Dl EP17
EP4 D4 D2 EP12
EP5 Dll D3 EP16
EP6 Dll D3 EP19
EP7 D5 D4 EP4
EP8 Dll D4 EP11
EP9 D10 D4 EP22
EP10 D13 D5 EP7
EP11 D4 D5 EP21
EP12 D2 D6 EP1
EP13 D8 D6 EP26
EP14 D7 D7 EP14
EP15 Dl D7 EP25
EP16 D3 D8 EP13
EP17 Dl D8 EP20
EP18 Dll D10 EP9
EP19 D3 Dll EP5
EP20 D8 Dll EP6
EP21 D5 Dll EP8
EP22 D4 Dll EP18
EP23 Dll Dll EP23
EP24 D14 D12 EP3
EP25 D7 D13 EP10
EP26 D6 D14 EP24
 -26-

References

Avery, C, C. Jolls, R. Posner, and A.E. Roth, 2001, "The Market for Federal Judicial Law
Clerks," University of Chicago Law Review Vol. 68, pp. 793-902.
Blum, Y., A.E. Roth, and U.G. Rothblum, 1997, "Vacancy Chains and Equilibration in
Senior-Level Labor Markets," Journal of Economic Theory Vol. 76, pp. 362-411.
Gale, D., 2001, "The Two-Sided Matching Problem: Origin, Development, and Current
Issues," International Game Theory Review Vol. 3, pp. 237-52.
, and L. Shapley, 1962, "College Admissions and the Stability of Marriage,"
American Mathematical Monthly Vol. 69, pp. 9-15.
Niederle, M., and A.E. Roth, 2003, "Unraveling Reduces Mobility in a Labor Market:
Gastroenterology With and Without a Centralized Match," Journal of Political
Economy Vol. 1 l l , pp. 1,342-52.
, 2004a, "Market Culture: How Norms Governing Exploding Offers Affect Market
Performance," NBER Working Paper wl0256, Cambridge, Massachusetts: NBER.
, 2004b, "The Gastroenterology Fellowship Match: How It Failed and Why It Could
Succeed Again," Gastroenterology Vol. 127, pp. 658-66.
Kagel, J.H., and A.E. Roth, 2000, "The Dynamics of Reorganization in Matching Markets: A
Laboratory Experiment Motivated by a Natural Experiment," Quarterly Journal of
Economics, Vol. 115, pp. 201-35.
Roth, A.E., 1984a, "The Evolution of the Labor Market for Medical Interns and Residents: A
Case Study in Game Theory," Journal of Political Economy Vol. 92, pp. 991-1,016.
, 1984b, "Misrepresentation and Stability in the Marriage Problem," Journal of
Economic Theory Vol. 34, pp. 383-87.
, 1984c, "Stability and Polarization of Interests in Job Matching," Econometrica Vol.
52, pp. 47-57.
, 1985, "The College Admissions Problem is not Equivalent to the Marriage
Problem," Journal of Economic Theory Vol. 36, pp. 277-88.
, 1986, "On the Allocation of Residents to Rural Hospitals: A General Property of
Two-Sided Matching Markets," Econometrica Vol. 54, pp. 425-27.
, 1989, "Two Sided Matching with Incomplete Information about Others'
Preferences," Games and Economic Behavior Vol. 1, pp. 191-209.
, 1990, "New Physicians: A Natural Experiment in Market Organization," Science
Vol. 250, pp. 1,524-28.
 -27-

, 1991, "A Natural Experiment in the Organization of Entry Level Labor Markets:
Regional Markets for New Physicians and Surgeons in the U.K.," American
Economic Review Vol. 81, pp. 415-40.
, 2002, "The Economist as Engineer: Game Theory, Experimental Economics and
Computation as Tools of Design Economics," Fisher Schultz lecture, Econometrica
Vol. 70, pp. 1,341-78.

, 2003, "The Origins, History, and Design of the Resident Match," Journal of the
American Medical Association Vol. 289, pp. 909-12.
Roth, A.E., and E. Peranson, 1999, "The Redesign of the Matching Market for American
Physicians: Some Engineering Aspects of Economic Design," American Economic
Review Vol. 89, pp. 748-80.
Roth, A.E., and U.G. Rothblum, 1999, "Truncation Strategies in Matching Markets: In
Search of Advice for Participants," Econometrica Vol. 67, pp. 21-43.
Roth, A.E., and M. Sotomayor, 1989, "The College Admissions Problem Revisited,"
Econometrica Vol. 57, pp. 559-70.
, 1990, "Two-Sided Matching: A Study in Game-Theoretic Modeling and Analysis,"
Econometric Society Monograph Series, Cambridge: Cambridge University Press.
, 1992, "Two-Sided Matching," in Handbook of Game Theory 1, ed. by R.J. Aumann
and S. Hart, Amsterdam: Elsevier, pp. 485-541.
Roth, A.E., and X. Xing, 1994, "Jumping the Gun: Imperfections and Institutions Related to
the Timing of Market Transactions," American Economic Review Vol. 84, pp.
992-1,044.

, 1997, "Turnaround Time and Bottlenecks in Market Clearing: Decentralized

Matching in the Market for Clinical Psychologists," Journal of Political Economy
Vol. 105, pp. 284-329.
Zhou, L., 1991, "Stable Matchings and Equilibrium Outcomes of the Gale-Shapley's
Algorithm for the Marriage Problem," Economics Letters Vol. 36, pp. 25-29.